Hydrogen and oxygen stable isotopa determinations have provided stimulating and fruitful results in the fleld of glaciology (e.g. see Hoefs, 1980, and Amason, 1981, for references). The seasonal variatlon of oD and o180 values of snow and fim can be used to date snow and fim layers as a function of depth. The isotopa ratios are set by temperature pl us the accumulation rate (mass of water preclpltated). For example, Epstein et al. ( 1965) used hydrogen and oxygen lsotopes to tind an average annua! accumulation rate of 7 cm of water at the South Pole between 1958 and 1963, In good agreement with stratigraphic and radioactlve datlng results.
Snow drift can affect the lsotopic compositlon of the snow after it has been deposited (Amason, 1981). The lsotopic ratios for the light isotopes are changed considerably every time there is a phase change (between vapor/llquld/solld), more at low temperatures than at higher temperatures, and as a function of mass ratlos lnvolved in the phase change reaction (e.g. Hoefs, 1980). Klnetic isotopa effects, common at low temperatures, may turther obscure the plcture as calculated by equlllbrtum Isotopa effects.
As a result, the seasonal fluctuations of oD and o180 in snow and ice are gradually eliminated due to homogenization processes, such as melting and refreezing of water percolating through snow, fim and lee. Thus, detailed comparisons between these isotopa records of the ice sheets in Greenland and Antarctica are unreliable (Hoefs, 1980). Amason (1981) glves examples of isotopa varlations in ice cores from the Byrd Station in Antarctica where It was Impossible to interpret successive summer and winter layers. The isotopic variation was toa lrregular to be acceptable for paleoclimatic use. Still the hydrogen and oxygen Isotopa values have been used as record of climatic conditions in terms of calculated mean air temperatures, like for the Vostok ice core (Jouzel et al., 1987; Bamola et al., 1987).
50
-For the Vostok ice core Jouzel et al. (1987) correlate the recent annual averages of the surface temperature with the recent snow hydrogen Isotopa ratio, uslng a correction model for the kinetlc isotopa effect. This relation Is then used to calculate temperature devlatlons from the present surface temperature (-55.5°C) for the whole 2083 m lee core covering 160,000 years. The authors must assume that the water preclpitatlon rate was constant throughout this lang time span covering several glacial and interglacial stages.
This assumption is highly questionable. Nor is It likely that the lee has been unattacked by phase changes and melt/lce-lnteractlons (see Chapter 5) that will alter the original light isotopa ratios.
Regarding the kinetic isotopa effects Jouzel et al. ( 1987) point out that "obviously, a change with time in the relative strength of this klnetlc effect, possibly associated with changes in saturation conditions, could partly obscure the relation between the isotopic content of polar precipitation and its [calculated] temperature of formation". They therefore conclude from their consideratlons that "deviations of up to 20% In the estimata of the temperature change cannot be excluded". Taking lnto account the conslderations discussed in Chapter 5. 1, it seems that the unfortunately highly uncertaln temperature calculations have been wldely adopted uncritically by non-scientists as well as scientists (e.g. Houghton and Woodwell, 1989), without taklng the uncertalnty or the systematics of the method lnto consideratlon.
7. LAG BETWEEN C02 LEVELS AND TEMPERATURE CHANGES
The noticeable warming tendency of the Northern Hemisphere surface air, whlch started around 1910 and reached a maxlmum in the late thirties, took place prior to the major "greenhouse gas" emissions (Michaels, 1990). In the Mauna Loa record of C02 atmospheric concentration (Keeling et al., 1989) the peak values lag behind the global continental temperature increases by about five months (Kuo et al., 1990). Kuo et al. (1990) noted that changing temperatures lead to changes In the amount of C02 outgassing or dissolution In the ocean, and to variation in biological activity, and thus to C02 Javel changes In the atmosphere. The five month lag time of C02 level behind temperature level changes, indicates that the causality Is: temperature-to-C02• lf C02 trapplng heat was the causa, an opposite sequence of maxima should be observed. Barnett (1990) stated that the
,
interrelation between the two varlables, i.e. co2 and temperature, may be due to temperature changes arislng from natura! cllmate variabllity, and has nothing to do with an increaslng "greenhouse effect".
The ocean is a large sink for C02 dlssolved in water. Because of C02's retrograde solubillty in water, lncreased temperature causes degasslng of C02 from the ocean to the atmosphere (Segalstad, 1990). Thls "thermal solubllity pump" accounts for 70% of the C02
flux from the ocean to the atmosphere, while "the blologic pump" accounts for the remaining • i 30% (Volk and Llu, 1988). It has been estimated that 4000 GT of C02 (equivalent to 1000 GT
of carbon) is fluxed from the ocean via the atmosphere to the continental biosphere when golng from a glacial to an lnterglaclal staga (Faure, 1990), owing to these two "C02 pumps"
only. Other natural carbon fluxes (weathering, volcanism, hot springs, carbonate sedimentation, degasslng by metamorphosis of rocks, El Nlflo - Southern Oscillation, etc.) are not lncluded in this flgure.
8. TEMPERATURE RECORDS
As lndicated before, it was assumed that since the middle of the 19th century the concentration of C02 increased in the atmosphere by .... 25%. According to the General Circulation Models (GCMs) thls should have increased the global mean air temperature by up to about 2.5°C today. The measurements do not confirm this prediction.
Attempts to tind a warmlng trend signal in the long temperature records are to a large extent based on lnspection of smoothed curves ("runnlng means") and not on more reliable time series analyses. The way In which smoothlng procedures may totally change the statistical propertles of the original time series, and lead to unrealistic conclusions, has been treated by several authors (sea e.g. Hisdal, 1956). Variations with long periods are emphasized in a fascinating manner, and short period "neise" is suppressed, the final result depending on the type of smoothing filter applied. Even completely random series may, on the basis of vlsual lnspection, show convincing "climatic changes" after having been beautified by such linear operations (cf. Haavelmo, 1951).
Hansen and Lebedeff (1987, 1988) found that the average global land temperature increased about 0.7°C during the past one hund red years. Other studies suggested that the·
global land and sea-surface temperatures rose by about 0.5°C (see review in Monastersky, 1989). Hansen (1988) stated in front of the U.S. House of Representatives that he had "99%
-52-confidence" in the reallty of the global warmlng trend, and In the cause/effect relationship between a 0.7°C global warming and anthropogenerated "greenhouse" alterations. Thls statement was criticlzed on theoretical grounds (see e.g. revlew in Kerr, 1989) and it was also found that the uncertalnty of the estimated past global temperatures is about the size of the warming signal (Barnett, 1989-after Monastersky, 1989).
A recent study of the NASA climate records in the United States showed a warming of 0.4°C in the twentieth century. However, this temperature lncrease could be strongly influenced by an urbanization effect on measurements taken at meteorological stations situated in or near the cities (Karl and Jones, 1989). Also a part of the increase turned out to be due to an error of the computer program when NASA supplied data to NOAA for analysis (Karl - after Michaels, 1990). A study of rural-statlon temperatures in the Soviet Union, China, and Australia suggests that the urbanization effect could be lower than expected (Jones et al., 1990).
Hanson et al. (1989) found that the records from the 48 contiguous United States do not indicate any statistically significant evidence of an overall lncrease In annual temperatures or change In annual preclpitatlon between 1895 and 1987.
A new study of the worldwide ocean temperatures ·slnce 1850, carried out by a group from MIT (Newell et al., 1989), shows llttle or no global warming over the past century. The authors found that the average ocean surface temperature Is now vlrtually the same as It was In the 19405. Jf two thlrds of the bulldup of C02 have taken place since 1940, the MIT data do not support the model predlctions.
Reynolds et al. (1989), analyzing surface and satellite measurements, concluded that there is no evldence of any warmlng trend in the ocean surface water between January 1982 and June 1988. This findlng refuted an earller clalm by Strong (1989) that the sea surface temperatures increased In thls period by as much as O.l oc per year.
The variability of natura! climatic fluctuations, and human influence on temperatures (urban and "heat Island" effects), have been described e.g. by Landsberg (1974) and Groveman & Landsberg (1979). These factors make it difficult to obtain homogeneous time series of annual global temperatures. During the winter, temperature variations over continents are, on the average, 4-6 times larger than over oceans, and 2-3 times larger on a yearly basis. The area ratio continents/oceans between 15 and 70° North is 0.88, but only 0.14 between O and 70° South. For an area weighed arithmetlc mean calculated for the Northem and Southem Hemlsphere, the temperatures measured over contlnents wlll dominate for the Northem Hemlsphere. A major problem Iles in maklng the weighed global
annual mean temperatures representative when more temperature data are available over continents than over oceans. This problem has still not been satisfactorily solved.
Long, relatively rellable series exist for the Northern Hemisphere from Europa, North America, Soviet, and Japan. These data were used by Borzenkova et al., (1976) and by Groveman and Landsberg (1979) for calculating weighted annual average temperatures for the Northern Hemisphere (Figure 7). In order to extend the time series back in time (between 300 - 400 years B.P.) Groveman and Landsberg used records of trea ring thickness from Finland and Alaska, which had correlated well with later instrumental temperature records. The older part of thelr time series is the least rellable one.
A spectral analysis of the almost 400 years long time series reveals that almost half of the variance can be explained by a perlodic variation with a wavelength of 99 years (Groveman and Landsberg, 1979). A possible explanation of thl� is cataclysmic volcanic eruptions occurring at about this time lnterval, e.g., Laki (lceland) 1783-1785 and Krakatau (Indonesia) 1883. These released sulfurdioxide exploslvely to the stratosphere, where the S02 combined with water to sulfuric acid, effectively blocking a part of the incoming solar radlation (Segalstad, 1983; Palais and Sigurdsson, 1989). It has been calculated from geologic evldence that Lakl's 1983 eruption caused the formatlon of some 90 million tons of sulfuric acid in the stratosphere, whlle the 1983 Krakatau eruption contributed some 30 million tons. Other volcanlc eruptions also contribute steadily to the atmospheric input of sulfur. For example the small 198 1 eruption of the Krafta volcano (lceland) contributed to the formation of some 100,000 tons of sulfuric acid (Palals and Sigurdsson, 1989).
Because of the lrregular lnfluence of volcanoes on the Earth's climate, It is important to look at time series which are several times longer than the perlodicity of about 100 years.
For a discussion on time spans and resolution of various climatic records see e.g. Webb (1985).
It is interesting to compare the almost 400 years long series with a shorter portion from the years 1850-1975. Both data sats show the same arithmetic means and standard de
viatlons, and are both normally distributed. Adapting a linear trend to the data, the long time series shows almost no change at all (+0.076°C per 100 years), white the short series (1850
- 1975) gives a small increase (+0.33°C per 100 years). It is important to note, however, that the calculated trends "explain" on ly about 15% of the respective variances. Jf one assumes that the current (and futura) lncrease in co2 releases from human buming of fossil fuels follows an exponentlal or logarithmic trend instead, and that thls trend would be seen in the
-
54-temperature record, one should note that ex
p
onential or logarithmic trends wlll not tit the data a ny better.Using statistical methods it is hardly possible to predict futura temperature changes.
It is necessary to tind the geophysical mechanisms behind climatic changes befare reliable predictions can be made.
The rescaled range methods of Mandelbrot and Wallis (1968, 1969 a, 1969 b) are appropriate for examination of certain long geophysical records. They were used, e.g., for studying the many thousand year long record of water level, flooding and drought of the river Nile. Such records have in most cases been found to be "fractal neise", and statistics based on random series cannot be used. This method tests the perslstence tendency of the data, i.e. if high or low levels tend to occur over a long time. It can also put constraints on the likelihood of getting extreme values in the futura.
Using this procedure Frøyland (1990) found for the almost 400 years long temperature series, in Figure 7, a streng tendency for the temperature to remain within the same range for times much lenger than what would be expected if there was statistical lndependence in the series. This means that there is a streng persistence tendency in the series, i.e. a warm year is likely to be followed by another warm year, a warm decade is likely to be followed by another warm decade, etc.
Based on examination of rather short (100-150 year) time series many researchers have concluded that a recent rise in temperatures is caused by only one, anthropogenic, causa. This signal is by the same researchers projected into the futura to give 1.5 to 4.5°C higher annual global temperatures 60 years from now.
By inspection of the long time series (Fig. 7) we sea three other similar temperat,Jre rises (starting at about 1835, 1675, 1605). lf It is true that the anthropogenic influence by buming of fossil fuels started about 1750, we should be able to see this as a signal in the temperature data over the last 200 years, and different from the foregoing 200 years. We see no obvious difference between these two halves of the series. Even a claimed recent ri�e in the 1980s {not shown in Figure 7) is less dramatic than the rise in temperatures from about 1812-1830.
In the past, excursions much larger than those in the 400 year record have taken place (Holocene warming, etc.). It is also well known from dynamic systems theory that phenomena like intermittent bursts can commonly occur in nonlinear systems.
o
c
15.0
14.6
14.2
13.8
13.4
1600 1650 1700 1750 1800 1850 1900 1950
OVI
o
250
500
750
Flgure 7. Lower curve: Northem Hemlsphere annual temperatures (data from Borzenkova
et al., 1976 and Groveman and Landsberg, 1979). Upper curve: Volcanic Dust Veil lndex,
O. V.l. (data from Lamb 1970, 1978), wlth lncreasing values downwards.
56
-We therefore conclude that the natura of the temperature data is such that no signal of human-induced global warmlng Is signlficantly traced in the best available temperature data for the Northem Hemisphere. High recent temperatures are not significantly different from earlier similar excursions (Fig. 7), and are not of a magnitude or duratlon (part of a 200 year long increasing trend) that is unique or alarmlng in the long-term (400 years) natural temperature fluctuatlons. Recent high temperatures are not different from what can be expected from statistical properties (dynamic systems theory).
The 3.3% annual rise in the release of C02 from buming of fossil fuels (EIIiott et al., 1985) since about 1750 should give a statistlcally signlficant signal in the temperature. It is very difficult to test the significance of a trend, because the temperature record represents a series with a streng persistence tendency. The temperature record may be characterized as resembling "fractal neise". From numerical analysis, the last 400 years temperature record is not showing characteristics that would make the prediction of 1.5 to 4.5°C global heating by the year 2050 very likely from a process that was acting in this period, regardless its causa.
9. AIR TEMPERATURES AND GLACIERS AT HIGH LATITUDES
Temperature data from the Scandinavian peninsula and Danmark for the last 120-130 years give no evidence of any increaslng trend during this period (Hanssen-Bauer, 1990).
There seems to be a slight lncreasing tendency up to the 1940s, followed by a slight decrease towards the end of the series. The warmest period In the 1930s coincided wlth the lowest volcanic dust loadlng of the global atmosphere during the last several hundred years
(Lamb, 1970) (see Figure 7).
Long-term temperature records in Svalbard started in 1912 at Green Harbour. During the first years the smoothed mean temperature rose considerably until the 1920s, particularly during the winter season (Birkeland, 1930; Hesselberg and Birkeland, 1940).
Since then there have been several "nice" waves in the Svalbard temperature records, as in all smoothed series, but no definite sign of an increasing "greenhouse" warming, that would justify a statistical analysls (Figure 8).
Five studies of other arctic temperature records (presented by Michaels, 1990) show similar variations as those found In the Norwegian Arctic.
MEAN TEMPERATURE
•c Jan. • March (5-year movlng average)
·8
-10
·12
·14
·16
·18
1920 1930 1940 1950 1960 1970 1980
Figure 8. Five-year movlng average temperature for Janua,Y-March for the statlons Jsijord Radio (78°04'N, 130J8'E) and Hopen (76°30'N, 25°04'E).
---
-- ..
58
-Using the malt layers at the Devon lee Cap in the Canadian Arctic, Koerner and Fisher (1990) reconstructed the summer Arctic temperatures during the last 10,000 years. Their data suggest that the warmest summers occurred 8,000 - 10,000 years ago, and the coldest ones only 150 years ago (end of "The Little lee Age"). The cooling from about 9,500 years ago to the present was estimated to be about 2.5°C.
Stratospheric temperatures, whlch according to climatic models should decrease In a
"greenhouse-gas" enriched atmosphere, did not show a statistically slgniflcant decllne since 1960 (Angell, 1986).
Because of the combined effect of the "greenhouse gases", we should, according to Michaels (1990), have effectlvely gone beyond half way to a doubling of C02• But the high latitude temperatures have simply not responded in the predlcted fashion. In fact the data indlcate a rise in temperature prior to the major "greenhouse gas" emissions, followed by a decline, or on a longer time scale, no trend at all.
Lefauconnier and Hagen ( 1990) and Hagen and Uestøl ( 1990) found at the Brøgger
breen and Lovenbreen glaciers in Svalbard an indication of a recent deceleration of loss of ice mass. This is in agreement wlth the study of Koerner et al. (1989) who have determined the ice mass balance for the past 10 to 30 years at four ice caps in the Canadian Arctic, and found no lndlcatlon of increaslng ablatlon. These lee caps and the Svalbard glaciers are likely to be good detectors of lncreasing "greenhouse effect", especially It It Is true that the warming should be most pronounced at high latltudes. In the Canadian ice cap Koemer et al. (1989) found that the condltlons during the last years constitute a marked contrast to the heavy melting years that characterized the warm period from the 1960s to the early 1980s. During the second half of the 1980s, there was an increase in the lee mass at the Melville South and Melghen lee caps. Since about 1968, the advance of small glaclers in West Green land coinclded wlth a period of decreasing summer temperatures, and could be seen as a direct responsa to this climatic deterioration. Six out of the nine glaciers studied contlnued to advance at least untll 1978 (Gordon, 1980).
One should also note that new .satellite :surveys indicate that both the lee caps of Green land and Antarctica are now lncreaslng, corresponding to a lowering of the sea water level of 0.45 and o. 75 mm per year, respectlvely (Meler, 1990). Hence the thermal expansion of the sea water during a warming period can thus be counterbalanced. Of course during cooling periods the water of the oceans wlll tend to show thermal contraction.
The above evldence supports the opinion (Wigley et al., 1989) that the anthropogenlc lncreaslng "greenhouse" warmlng signal has not yet been detected In a rigorous way.
1 O. CONCLUDING REMARKS
* The atmosphere is a contemporary, rather short-term, storage for the trace gas C02, In which this gas has a residence time of about 5 years.
* The atmospheric C02 ls constantly changlng and adjustlng its concentratlon according to the natural changes In the Earth's temperature. This is governed by inorganic thermodynamic gaseous, aqueous and mineral equlllbria, and by blologic processes.
The anthropogenlc C02 is negllgibly small compared to these streng C02 "pumps" and other natural fluxes of co2.
* C02 has a high solubllity in water. Lower aqueous solu_bility of C02 at higher temperature wlll make the oceans degas C02 to the atmosphere, when the sea and
* C02 has a high solubllity in water. Lower aqueous solu_bility of C02 at higher temperature wlll make the oceans degas C02 to the atmosphere, when the sea and